Innovation Profs - 1/30/2024

Your weekly guide to generative AI tools and news

Author

Innovation Profs
January 30, 2024

Generative AI News

Microsoft adds new restrictions to Designer AI used to make Taylor Swift deepfakes

You’ve likely heard about the explicit deepfakes of Taylor Swift that were recently circulating on the internet on sites like X (née Twitter) and Reddit. The source of these images has been traced to Microsoft’s Designer AI, powered by image generation model DALL-E 3. Now Microsoft is adding new restrictions to the tool, as users apparently were able to bypass safeguards that were put in place to prevent the production of images of celebrities in explicit situations.

FTC opens inquiry into Big Tech's partnerships with leading AI startups

The US Federal Trade Commission has opened antitrust investigations into the relationships between tech giants Amazon, Google, and Microsoft and AI startups Anthropic and OpenAI. As we’ve reported, Microsoft owns a 49% stake in OpenAI, while Amazon and Google have invested billions in Anthropic. According to FTC chair Lina Khan, “We’re scrutinizing whether these ties enable dominant firms to exert undue influence or gain privileged access in ways that could undermine fair competition.”

Anthropic confirms it suffered a data leak

Last week the AI company Anthropic announced a data leak, as a contractor unintentionally sent a file with user data to a third party. No sensitive user data was included in the file. Anthropic further characterized the leak as an “isolated error.” Still, the announcement comes at a bad time for Anthropic, who are one of the companies included in the FTC probe described above.

Google’s Hugging Face deal puts ‘supercomputer’ power behind open-source AI

Hugging Face, an AI model repository that hosts over 350,000 open-sourced models, has formed a partnership with Google Cloud, which will allow Hugging Face users to “build, train, and deploy AI models without needing to pay for a Google Cloud subscription.” In particular, users will have “cost-effective” access to Google’s vast trove of TPUs (tensor processing units) and GPUs (graphics processing units) that are used for training machine learning models.

iOS 17.4: Apple continues work on AI-powered Siri and Messages features, with help from ChatGPT

We still haven’t seen Apple enter the generative AI rat race, but it is now expected that Apple will incorporate generative AI into iOS 18, which will likely be announced in June. In particular, Apple is reportedly developing an LLM-powered version of Siri and will also incorporate generative AI into Messages. As reported by 9to5Mac, code found in the beta version of iOS 17.4 includes a feature called SiriSummarization, which includes calls to OpenAI’s ChatGPT API (which is likely used to test the performance of Apple’s models against ChatGPT and is not likely to be part of iOS 18).

Quick Hits

Study finds AI ‘revolution’ moving at a crawl in enterprises

A new survey reveals that despite the hype around generative AI, real-world enterprise adoption remains low due to barriers like skills gaps, infrastructure constraints, and overall AI integration challenges.

venturebeat.com/ai/study-finds-ai-revolution-moving-at-a-crawl-in-enterprises

We Asked A.I. to Create the Joker. It Generated a Copyrighted Image.

Artists and researchers are exposing copyrighted material hidden within A.I. tools, raising fresh legal questions.

www.nytimes.com/interactive/2024/01/25/business/ai-image-generators-openai-microsoft-midjourney-copyright.html

Midjourney's V6 update brings Pan, Zoom, Vary and broader website access

Midjourney has added Pan, Zoom and Vary (Region) to its V6. The pan feature now acts more like zoom, promising higher quality results with "more coherence and less repetition of concepts".

the-decoder.com/midjourneys-v6-update-brings-pan-zoom-vary-and-broader-website-access

How enterprises are using open source LLMs: 16 examples

Open-source advocates agree there are many more examples of closed-model deployments, but it’s only a matter of time before open-source catches up with the closed-source models. 

venturebeat.com/ai/how-enterprises-are-using-open-source-llms-16-examples

New Theory Suggests Chatbots Can Understand Text | Quanta Magazine

Far from being “stochastic parrots,” the biggest large language models seem to learn enough skills to understand the words they’re processing.

www.quantamagazine.org/new-theory-suggests-chatbots-can-understand-text-20240122

Can AI do ugly? | TechCrunch

Some thoughts on AI aesthetics, the challenge of uninsurability, and how to pitch a biotech startup to non-experts.

techcrunch.com/2024/01/27/ai-versus-reality/?guccounter=1&guce_referrer=aHR0cHM6Ly9uZXdzLmdvb2dsZS5jb20v&guce_referrer_sig=AQAAAKrY6lqJIsaW3vJ9NaF36FYPb3dDJKdv9vsBoFBUh1P60uO7xy33h65PLGhOd9QndVyKc_J3Tec_B5NejZCVhrvHPPoI2NfAd_EyShWbAIdBKKPyG79K8ZWHJLvntP3BZ9-4S6UbLj3GHRCXnIorpbb-O6ZSbIu9eYLxcKOCDC41

Generative AI’s enterprise gamble: IT leaders bet big on tech despite security woes

A new report finds that enterprise IT teams are rapidly increasing budgets and adopting generative AI despite risks, but many lack clear ROI metrics and are willing to implement the technology quickly even if negative consequences could occur.

venturebeat.com/ai/generative-ais-enterprise-gamble-it-leaders-bet-big-on-tech-despite-security-woes

Tool of the week: Lumiere

Google showed off its AI video generator Lumiere last week. The tool can make five-second videos from text and from images. It can also base videos off of the style of another image or video.

Lumiere is not yet available to the public, but you can preview it here.

If you are looking for a tool that makes generative AI videos now, try Runway or Pika.

AI-generated image of the week

This was an image we made in our Intro to Generative AI workshop last week from a Mad Libs-style prompting exercise. Which of the four do you like best? 

Prompt: an epic teal teenager surfing in a cornfield using the colors purple and yellow in the style of oil painting

Generative AI tip of the week

Here are elements to include in an image prompt (see them represented in the image prompt above):

  • Main subject

  • Secondary subject

  • Artist names or styles

  • Photography prompts

  • Specific colors

Get starting with Generative AI

New to generative AI? Here are some places to start…

What we found

Artificial World on X used Midjourney to make “Grant Theft Auto” versions of famous people, including Joe Biden, Donald Trump and Abe Lincoln.